SaaS is dead, and AI killed it.

Or that’s what several loud voices in the tech industry would have you believe, anyway. And yes, there’s a grain of truth in the statement: the rise of vibe coding and AI systems means that even those with limited technical knowledge can consider ditching fancy — not to mention expensive — dashboards and replace SaaS components with automated agentic solutions.

Vercel’s CEO, Guillermo Rauch, sparked a lot of conversation (and panic) when he posted: “Almost every SaaS app inside Vercel has now been replaced with a generated app or agent interface, deployed on Vercel.” He continues that “we’ve also been moving off legacy systems with poor, slow, outdated, and inconsistent APIs, as well as just dropping abstraction down to more traditional databases.”

But there’s more going on under the surface than that suggests. SaaS users don’t just pay for a shiny UI — they pay for convenience, security, ease of use, service-level agreements (SLAs), ongoing maintenance, and customer service. Which begs the question: are agents truly capable of replicating the traditional SaaS experience yet?

Software consumption habits are changing in the age of AI, which poses an existential threat to SaaS providers. The UK government’s National Centre for Cyber Security has publicly floated the idea that AI might eventually kill SaaS as we know it. But we’re not there yet.

As we’ll see below, robust APIs (as opposed to fancy dashboards) are a vital ingredient in this shift, acting as an execution layer for agentic AI. And SaaS providers with well-designed APIs will be far better positioned in this new era.

From Dashboards to Autonomous Agents

Historically, people have bought (or rented, since you never truly “own” a SaaS product) for all sorts of reasons. Alongside lower upfront costs and high scalability, SaaS exposes functionality across organizations in a way that one-time licenses for costly enterprise products likely can’t.

Cloud-based SaaS platforms store business data and workflows, enabling anyone in the organization to access them, with permissions and governance handled inside the product. AI agents can’t yet, at least not without significant customization, replace all of that. The idea that agents are replacing SaaS is a fallacy, because what they’re really altering is the point of consumption.

Where SaaS might require you to set up a workflow manually, by configuring integrations and importing data, an AI agent that’s connected in the right way can be prompted conversationally. This might look like: “Match sales data with recurring customers, and identify those who have not purchased in more than three months so we can email them a discount code.”

That’s a powerful concept because it changes the focus from navigating a UI to using a system to execute natural language commands. But to work effectively, such a system requires structured access, permission-awareness, authentication, and orchestration across various platforms, and event-driven updates. In other words, it requires well-documented, low-latency, reliable APIs.

APIs Become More Valuable (and Expensive?)

OpenAI, Anthropic, and Google (as well as others) all ran ads focused on AI during the Super Bowl. It’s perhaps the best indicator, in case it wasn’t already clear, that AI companies have mass adoption in mind. There’s no doubt that they’re already looking to erode the competitive advantage that user-friendly SaaS products currently enjoy.

One thing remains certain: whether they’re being consumed via chained agentic workflows or more traditional integrations via SaaS, APIs are becoming more important than ever. They’re the battleground on which this struggle is playing out and are, as a result, like gold dust.

We’ve previously written about the reasons public APIs are shuttering in the age of AI and one of them is that, presented with a new opportunity to price their products based on the type and scale of use that agentic consumption facilitates, some providers are choosing to restrict access and create their own AI-powered services instead. We might see much more of that soon.

As it stands, AI-powered tools designed to sidestep SaaS should be treated with caution when scaling up. When one link in a chained workflow breaks due to, say, a shuttered API or a deprecated service, the entire product fails. If that data proves to be irreplaceable, the person who built that product is back to square one. For that reason, this process is extremely fragile.

SaaS Companies Must Evolve to Survive

Despite the existential risks, SaaS tools retain advantages that they can lean on to remain indispensable. Perhaps the biggest of these is security, with reputable SaaS providers taking many more precautions around safety and uptime than a vibe coder merrily feeding private customer data into poorly secured large language models (LLMs). Another is reduced key man risk.

SaaS solutions are generally regarded as heavily democratizing, granting access to top-tier enterprise tools at a reduced cost, often via a graphical or low-code/no-code UI. While the average business user can figure out how to use a SaaS product, that isn’t yet true of chaining tasks together using AI agents, LLMs, or Model Context Protocol (MCP). Teams that lean into replacing SaaS with AI risk creating dependencies and bottlenecks around those who have the know-how.

It’s not surprising that we’ve seen tools like Zendesk, Salesforce, and Notion heavily promote the integration of AI into their services. Part of the reasoning may be to limit customers from jumping ship to build their own custom agentic solutions. Meanwhile, we’ve also seen automation tools like Zapier and Workato ramp up their use of agents, workflows, and MCP. But is wrapping up AI in a shiny UI enough for SaaS to survive?

SaaS Isn’t Dead. Its Consumption Layer Is Changing

Historically, SaaS has been successful because it replaces the complex processes associated with software development — building, deploying, and maintaining infrastructure — and packages it up as something accessible, affordable, scalable, and, yes, often pretty to look at.

Although the rise of AI agents suggests that the pendulum is swinging the other way, with its emphasis on prompting, the shift towards command-based interaction isn’t quite as cut and dry as it might seem. Consider the default greetings of ChatGPT: not “Enter prompt here,” but warm welcomes like “What are you working on?” and “What’s on your mind today?”

Likewise, many AI companies and AI-enhanced tools have quirky characters, such as Salesforce’s Astro, Claude Code’s lobster (Clawd), and the Skales lizard. Mascots alone don’t warrant the cost of a monthly subscription, however cute they might be, but they’re a microcosm of user appetite for visual abstraction: dashboards, reporting views, presentation layers, and the like.

A finance team might be able to make sense of a budget forecast generated by an agent, but they’ll probably want some polished visuals to show the CEO. That’s something most SaaS products are very good at generating. Auditability, observability, executive reporting, and human approval layers are areas in which UI can make a case for itself to persist in enterprise software.

MCP and agents may not replace SaaS products so much as reshape how these products work. Accio Work, a business automation platform that runs entirely on your machine but uses LLM requests and external APIs for orchestration, is an interesting example of what the future might look like. It’s effectively AI-powered automation-as-a-service, or AaaS if you’re brave enough for another acronym.

Crucially, however, services like Accio Work don’t aim to replace SaaS. It may offer Shopify Operator and CRM agents, but you still need accounts on services like Shopify and Odoo for those agents to connect with. As such, the SaaS companies best positioned for the next wave of software consumption aren’t the ones with the fanciest dashboards, but the most robust APIs.

There’s a real possibility that we’re headed towards a headless world in which the UI is optional. The dashboard is no longer the product, but access to data is, which is something that many SaaS providers have been accumulating for years, sometimes even decades. Effective APIs continue to be the go-to mechanism for exposing data safely and securely, whether to humans or agents.

Which means that products with weak APIs risk becoming… well, invisible.